Elixir-AI/french-kid-positive-only

法国关键信息文件（KID）数据集包含财务文件（主要是KID）的图像，以及由Gemini 2.0 Flash Lite模型生成的与图像内容相关的财务问题和答案。数据集还包括文档的语言（以ISO双字母代码表示）和页码。该数据集旨在用于训练专门分析财务文件的视觉-语言模型（VLM）。

The French KID (Key Information Document) Dataset contains images of financial documents (mainly KIDs) along with finance-related questions and answers generated by the Gemini 2.0 Flash Lite model based on the image content. The dataset also includes the language of the document (in ISO two-letter code) and the page number. This dataset is designed for training vision-language models (VLM) specialized in the analysis of financial documents.